The Brazilian Connection in Milton Friedman’s 1967 Presidential Address and 1976 Nobel Lecture Mauro Boianovsky 1. Some Crucial Monetary Episodes from Brazil As has been often pointed out, the criticism put forward by Milton Fried- man (1968) and Edmund Phelps (1967, 1968) of the stability of the Phillips curve was mainly theoretical, not empirical (see Nelson 2020: 165–66, and references cited therein). Friedman (1977a: 455) acknowledged as much in his Nobel Memorial Lecture. Phelps and Friedman were not aim- ing to elucidate a riddle posed by empirical evidence. Estimates of the relation between the rate of change in nominal wages (or prices) and unemployment performed relatively well in the 1960s, largely duplicating the nonlinear inverse “successful” (Phelps 1968: 680) fitting of British I benefited from helpful comments by Edward Nelson, Joaquim Andrade, Alexandre Andrada, Antonio Delfim Netto, Edmar Bacha, David Laidler, G. C. Harcourt, Carlos G. Langoni, Bruna Ingrao, Guillermo Calvo, James9, André Lara Resende, Charles Goodhart, Rogério Arthmar, André Villela, Pedro Duarte, Ramón Fernandez, Jorge Soromenho, Gilberto Lima, Ana M. Bianchi, Mauricio Coutinho, Fernando H. Barbosa, Aloisio Araújo, Pedro C. Ferreira, Sylvie Rivot, Matthieu Renault, Marco Cavalieri, Carolyn Biltoft and (other) participants at seminars at Universidade de São Paulo (August 2018), FGV-Rio (October 2018) and Universidade Fed- eral do Paraná (April 2019), at the Brazilian Economic Meetings (Anpec, Rio, December 2018) and at the European Society for the History of Economics Conference (Lille, May 2019). Com- ments and suggestions by two anonymous referees and HOPE’s editor Kevin Hoover helped to improve the quality of the paper. I would like to thank the Hoover Institution and CPDOC-FGV for permission to quote from the Milton Friedman Papers, the Eugenio Gudin Papers, and the Paulo Egydio Martins Papers, and Elio Gaspari for providing information from the Golbery do Couto e Silva Private Papers. Research funding from CNPq is gratefully acknowledged. History of Political Economy 52:2 DOI 10.1215/00182702-8173418 Copyright 2020 by Duke University Press Downloaded from https://read.dukeupress.edu/hope/article-pdf/doi/10.1215/00182702-8173418/740664/0520367.pdf by [email protected] on 06 February 2020 368 History of Political Economy 52:2 (2020) historical data by A. W. Phillips (1958). Indeed, a main purpose of Fried- man and Phelps was to “reconcile” the empirical Phillips curve with the theoretical “neutrality axiom of anticipated inflation,” as put by Phelps (1968: 682). They did that by postulating that the Phillips curve shifts upwards (downwards) with the expected rate of increase (decrease) of prices (or wages), accompanied by convergence of actual unemployment to its “natural” or equilibrium level. However, Friedman (1968: 8–9) did use a piece of empirical evidence in support of his hypothesis about the role of expectations of inflation in determining the position of the (short-run) Phillips curve. He referred to the 1964–66 Brazilian stabilization plan and its perverse effect on employ- ment due to persistent anticipations of inflation. Friedman’s (1968) discus- sion of inflation and unemployment in Brazil is restricted to just one para- graph. But it is a crucial one, as that is also the only paragraph in his 1967 Presidential Address in which Friedman discussed Phillips (1958) and charged him for failing to distinguish between nominal and real wages. The effect of expectations of inflation on Brazilian money wages provided a forceful illustration of the instability of the Phillips curve, he claimed. Friedman (1966: 59) had referred to the Brazilian episode as “the most dramatic example” of the “fallacy” that there is a lasting trade-off between inflation and employment. In the first draft of his 1967 address, Friedman (1967a: 22) wrote that, for relatively low inflation rates as in the US and the UK at the time, it was easy to interpret the time series as “a trade-off between price-rise and unemployment rather than between acceleration of price-rise and unemployment.” Hence, “to distinguish between these, one must look at a broader range of experience. The difference is then patent.” It is implicit that the monetary history of countries beset by chronic high inflation, like Brazil and other South American nations, should be able to provide the required evidence. Friedman was primarily an empirical economist who used general eco- nomic principles to scrutinize facts in order to make predictions. As argued by Hirsch and de Marchi (1990), a close reading of Friedman’s (1953) methodological essay in the context of his economics indicates that he was much closer to the pragmatism of John Dewey than to the falsifica- tionism of Karl Popper. Instead of the Popperian separation between the contexts of discovery and justification of hypotheses, Friedman was con- cerned with the process of inquiry. From that perspective, theories follow as results of acquaintance with facts, in the sense that empirical investiga- tion is prominent not just in assessing a theory but also in developing it. The empirical basis of theories comes from working back from fact obser- Downloaded from https://read.dukeupress.edu/hope/article-pdf/doi/10.1215/00182702-8173418/740664/0520367.pdf by [email protected] on 06 February 2020 Boianovsky / The Brazilian Connection 369 vations to assumptions. He found much inspiration for his economic methodology in Alfred Marshall and Wesley C. Mitchell (see also Back- house 1996, chap. 11; Hoover 2009). Friedman’s work as a monetary economist was dominated by his study of money and the business cycle (with Anna Schwartz) for the National Bureau of Economic Research (NBER). The highpoint of that project— started in the late 1940s and continued for more than three decades, until the early 1980s—was their 1963 Monetary History (see Hirsch and de Marchi 1990, chap. 10; Hammond 1996). That book epitomized Fried- man’s methodology, with its stress on historical episodes and rejection of conclusive statistical tests as decisive for choosing among alternative hypotheses.1 Friedman’s case for the role of money as a prime determi- nant of economic fluctuations was built on the investigation of selected crucial episodes, such as the contraction of money supply in the US Great Depression in the early 1930s, mentioned in his 1967 address (Friedman 1968: 3). Indeed, he would often prefer the historical “episodic approach” to a full statistical treatment (Friedman and Schwartz 1982: 432; see also Friedman 2005). Economic history and economic theory are, therefore, intertwined in Friedman’s economics. As put by Hirsch and de Marchi (1990: 244), the methodological posi- tion that came out of Friedman’s work as a monetary economist was that a “rich appreciation of what is to be explained is crucial. Dramatic and discriminating tests have almost no place; the marshaling of evidence to strengthen conviction is almost everything.” That is consistent with Fried- man’s 1953 argument that a fruitful hypothesis is: “a way of looking at or interpreting or organizing the evidence that will reveal superficially dis- connected and diverse phenomena to be manifestations of a more funda- mental and relatively simple structure” (Friedman 1953: 33). Friedman’s 1967 presidential address has been seen as an exception to the primacy of empirical evidence in his monetary economics. Backhouse (2007: 17), for instance, has mentioned the “irony” that Friedman’s point about the accelerationist Phillips curve arose out of purely theoretical arguments with no empirical ground.2 In the same vein, Blanchard (2017: 1. As Friedman recollected in a November 2, 1993, letter to Robert Leeson, “I felt very con- fident [in my 1963 joint book] in the evidence from history independently of the evidence from the statistical correlation, and hence regarded these as confirmatory rather than decisive evi- dence” (Leeson 1998: 76). 2. However, Friedman 1968 may be read as offering a largely deductive argument about some crucial implications of the quantity theory of money, a theory he had extensively investi- gated from an empirical perspective in Friedman and Schwartz 1963 and other pieces. Downloaded from https://read.dukeupress.edu/hope/article-pdf/doi/10.1215/00182702-8173418/740664/0520367.pdf by [email protected] on 06 February 2020 370 History of Political Economy 52:2 (2020) 164) has described the development of the natural rate hypothesis by Friedman and Phelps in 1967–68 as “theory ahead of facts.” It was only in the 1970s, with the recorded coexistence of rising inflation and unemploy- ment—as well as some evidence of a vertical long-run Phillips curve—in industrialized countries, that Friedman’s natural rate hypothesis gained ascent (see also Forder and Sømme forthcoming). Nevertheless, as dis- cussed below, the Brazilian monetary experiment, as seen by Friedman, represented a “crucial episode” that provided a forceful illustration of what was to be explained. Friedman was very likely briefed about Brazil by other Chicago econo- mists with knowledge of South American economies, particularly Arnold Harberger and Larry Sjaastad, who had been frequent visitors to the region since the 1950s (see Harberger 1997) and the 1960s in the case of Sjaastad. In 1965, Harberger created and became the first director of the Center for Latin American Economic Studies at the Chicago economics department. Even before that, Latin American economies had attracted the attention of Chicago development economists, led by Theodore W. Schultz (see e.g., Schultz 1956). Friedman (1977a: 464) would refer to Harberger 1966 and Staajstad 1974 as sources about chronic inflation in South America. Harberger took part in the 1963 Rio conference about inflation and growth in Latin America, a major international event spon- sored by the Economic Growth Center of Yale University (Baer and Ker- stenetzky 1964; see Harberger 1964, based on his better known 1963 essay about Chile). Friedman did not participate, but probably knew about the conference volume, reviewed in the Economic Journal, the American Economic Review, and other journals.